JP7232192B2 - Data-Driven Respiratory Body Motion Estimation Method - Google Patents

Description

The following generally relates to medical imaging techniques, radiometric imaging techniques, positron emission tomography (PET) imaging techniques, single photon emission computed tomography (SPECT) imaging techniques, patient monitoring techniques, and related techniques.

In radiometric imaging, such as positron emission tomography (PET) or single photon emission computed tomography (SPECT), a patient or other imaging subject may be given an image that preferentially accumulates in target organs or tissues, and/or glucose absorption, etc. radiopharmaceuticals designed to track the bodily functions of Radiopharmaceuticals include radioactive isotopes, such as positron emitting isotopes of PET. The imaging subject is placed in an imaging device (eg, a PET scanner for PET imaging or a gamma camera for SPECT imaging) and radiographic imaging data is collected and reconstructed. Typically, analytical or iterative reconstruction algorithms are used to generate reconstructed images. To improve accuracy, an attenuation map of the imaged subject's body is provided, calculated, for example, from transmission X-ray computed tomography (CT) images of the subject, and the attenuation map is used to detect Attenuation of the emitted radiation (eg 511 keV gamma rays for PET) corrects the degradation of the reconstructed image.

A further cause of image degradation is subject movement. Subject inhibition can be used to suppress voluntary movement. However, unintentional body movements, especially breathing and, to a lesser extent, cardiac cycle movements are unavoidable. Radiopharmaceutical doses are kept small in order to limit radiation exposure to the imaging subject, and therefore typically take from a few minutes to collect sufficient data for an acceptable signal-to-noise ratio (SNR). A radiation imaging data acquisition time extended to several tens of minutes is used. The acquisition time therefore spans multiple respiratory and cardiac cycles.

Respiratory gating is a known technique for limiting the impact of respiratory motion on radiation imaging quality. In this approach, a respiratory monitoring belt or other respiratory monitoring device is attached to the imaged subject during radiographic data acquisition. The radiation data is binned (sorted) according to a gating scheme, and often only radiation data with a desired gating interval (indicated by the gating scheme) is reconstructed to produce a reconstructed image. Normally, this period from the completion of expiration in one breath to the beginning of inspiration in the next breath is static and usually occupies a significant portion of the respiratory cycle, e.g., about 60% in some cases. As it continues, the end-expiratory stage is chosen for reconstitution. Cardiac gating is performed in a similar manner, eg, using an electrocardiograph (ECG) as a cardiac cycle monitor. However, in cardiac gating studies, the radiation data at each gating step is reconstructed to form a gated cardiac image. Multiple gated cardiac images are then displayed in cine mode to show the motion of the beating heart.

New and improved apparatus and methods are disclosed below.

In one disclosed aspect, a radiographic imaging data processing apparatus comprises an electronic processor and a non-transitory storage medium storing instructions readable and executable by the electronic processor to perform a respiratory body motion estimation method. The method is to reconstruct radiographic data to produce a reconstructed image, where the radiographic data are lines of response (LOR) acquired by a Positron Emission Tomography (PET) imager. ray) or projections acquired by a gamma camera; using the reconstructed image to determine a motion assessment volume over a group of voxels; and for each time interval bin, the LOR or projection ( and generating a curve of displacement versus time by an operation including calculating a statistical displacement metric. Determining the motion assessment volume using the reconstructed image includes identifying features of the motion assessment image in the reconstructed image and creating a motion assessment volume that overlaps or includes features of the motion assessment image. and selecting.

In another disclosed aspect, a non-transitory storage medium stores instructions readable and executable by an electronic processor for performing a respiratory body motion estimation method by processing operations, the processing operations comprising: and generating a reconstructed image, wherein the radiographic data includes LOR acquired by a positron emission tomography (PET) imager or projections acquired by a gamma camera; , identifying a plurality of candidate image features in the reconstructed image; computing a statistical measure for each candidate image feature; Selecting a motion assessment image feature from a plurality of candidate image features based on a metric, and calculating a displacement vs. time representation of motion of the motion assessment image feature from the acquired radial list mode data. and generating a curve.

In another disclosed aspect, the displacement versus time curve provides a respiratory motion signal that can be used to define the range, number, and magnitude of amplitude-based respiratory gates. An amplitude-based respiratory gating scheme is used to acquire respiratory-gated reconstructed images at various respiratory stages.

In another disclosed aspect, a radiographic data processing method is disclosed. Amplitude-based respiratory gating schemes directly relate to the amount of motion blur within each respiratory gate or respiratory stage. These allow for more accurate motion compensation and effective motion blur reduction in respiratory motion compensation methods.

In another disclosed aspect, amplitude-based respiratory gating is applied to respiratory motion compensation schemes to reduce respiratory motion blur in reconstructed images. Acquired radial list mode data is binned into respiratory gates of amplitude-based equal count bins. These are reconstructed from the LORs or projections within each respiratory gate to form a series of amplitude-based respiratory-gated images. A series of motion vector images are computed between adjacent breath-gated images representing motion of features in the motion assessment images. The motion vector image is then used to transform the respiratory gated image into a respiratory motion-free image at a single reference stage, free of respiratory motion blurring effects.

In another disclosed aspect, a radiographic data processing method is disclosed. The radiation imaging data is reconstructed to generate a reconstructed image. Emission imaging data includes LOR acquired by a Positron Emission Tomography (PET) imager or projections acquired by a gamma camera. Features of the motion assessment image are identified in the reconstructed image. The radiographic data is binned into time interval bins based on LOR or projection timestamps. A displacement versus time curve representing the motion of the motion-assessed image feature is generated by calculating the displacement metric of the motion-assessed image feature for each time interval bin.

One advantage is to provide respiratory-gated radiation imaging without the use of a respiratory monitor and with improved signal-to-noise ratio (SNR) compared to other gating techniques that do not utilize a respiratory monitor. That's what it is.

Another advantage resides in providing respiration-gated radiation imaging without the use of a respiration monitor, which can be used for various types of radiopharmaceuticals that are variably distributed within a subject's organs and/or tissues. can be adapted to

Another advantage resides in providing respiration-gated radiation imaging without the use of a respiration monitor, which captures motion of organs or other image features that are not completely within the image field of view. It is available.

Another advantage resides in providing respiratory gated radiation imaging without the use of a respiratory monitor, which is computationally efficient due to the reduced range of image reconstruction required in generating the gating signal. It is good.

Another advantage resides in providing respiration-gated radiometric imaging without the use of a respiration monitor, which is computationally more efficient than using acquired list-mode data directly. be.

Another advantage resides in an amplitude-based respiratory gating method that relates the amplitude of respiratory motion to the amount of image blur. As a result, amplitude-based respiratory motion correction is not the widely used time-based or step-based method because respiratory motion is irregular both in terms of duration and amplitude of breaths from beat to beat. is more effective than the respiratory motion compensation method of

The presented embodiments provide some of the aforementioned advantages and/or other advantages that will become apparent to those of ordinary skill in the art upon reading and understanding this disclosure.

The invention may take form in various components and arrangements of components, and in various steps and arrangements of steps. The drawings are for purposes of illustrating preferred embodiments only and should not be construed as limiting the invention.

1 schematically illustrates a radiation imaging system with respiratory gating; FIG. 2 is a schematic flow diagram of an exemplary process for performing respiratory motion estimation for the radiation imaging system of FIG. 1; 2 is a schematic flow diagram of an exemplary process for performing respiratory motion estimation for the radiation imaging system of FIG. 1; FIG. 11 schematically illustrates a preferred motion assessment volume overlying the lung/thoracic diaphragm boundary; 5 schematically illustrates the generation of a one-dimensional (1D) displacement versus time curve by calculating total counts of radiographic data within the motion assessment volume of FIG. 4 for successive time interval bins; FIG. . FIG. 11 schematically illustrates another suitable motion assessment volume that overlaps or encompasses the lung/thoracic diaphragm boundary; 2 is a schematic flow diagram of an exemplary process for performing respiratory-gated image reconstruction of the radiation imaging system of FIG. 1;

Respiratory (e.g., breathing) motion estimation techniques disclosed herein rely on quantifying the motion of organs, anatomical boundaries, or other image features that serve as motion assessment image features. Operate. Respiratory motion is quantified by evaluating motion features in a motion assessment image of the acquired radiographic data. To identify motion assessment features, the acquired radiographic data is reconstructed to produce a reconstructed image. The reconstructed image then contains some degree of blur due to the respiratory motion of the imaged subject during the acquisition of the radiographic data, but is sufficient to be used to identify features for image evaluation. be. Since the reconstructed image is only used to identify features in the motion assessment image, it may be done without performing attenuation correction using an attenuation map. This speeds up the reconstruction process, increases counts, and makes features more detectable. For reconstruction, reconstruction using a voxel size larger than that used to reconstruct the clinical image, and/or using analytical or non-iterative image reconstruction algorithms, such as filtered backprojection. , other computationally efficient techniques can be used.

As used herein, the optimal organ, anatomical boundary, or other image feature of interest for estimating respiratory motion is the dose and type of radiopharmaceutical administered to the imaging subject, the radiopharmaceutical It depends on a number of factors, such as the delay period between drug administration and radiographic data acquisition, subject-specific factors such as body weight, girth, individual anatomy and/or metabolism, and the selected field of view (FOV). recognized as different depending on With this in mind, some exemplary respiratory motion estimation techniques disclosed herein provide multiple candidate image features in a reconstructed image (blurred by respiratory motion). , such as the heart, liver, and lungs in one illustrative example. This feature can also be an abnormality, such as a lesion. A statistical measure is computed for each candidate image feature, such as the average image intensity of the candidate image feature and/or the maximum image intensity gradient of the candidate image feature in the reconstructed image. The features of the motion assessment image are then selected from the plurality of candidate image features based on the statistical measures computed for the candidate image features. For example, this selection is based on the candidate with the highest average image intensity, since average image intensity implies substantial data acquired for the image feature of the candidate that makes the estimation of respiratory motion more accurate. select image features of

In further exemplary embodiments, centroids of candidate image features and their positions within a selected three-dimensional (3D) volume of interest (VOI) within the reconstructed image, and/or A statistical measure, such as counts within the volume of interest as the feature enters and exits the volume of interest, is computed for each feature of the candidate image. The features of the motion assessment image are then selected from the plurality of candidate image features based on the statistical measures computed for the candidate image features. For example, a candidate image feature is the location of the heart's centroid within a volume of interest that encompasses the entire heart during respiratory motion. This information would then be used to extract the respiratory motion signal. In another example, the candidate image feature is a sum of constant volume of interest counts covering the upper part of the liver, such that the upper part of the liver moves in and out of the volume of interest during respiratory motion.

In one embodiment, after selecting features for the motion assessment image, the radiographic data is binned into consecutive time interval bins based on LOR (for PET imaging) or projection (for SPECT imaging) timestamps. be done. For example, in one embodiment, after selection of a motion assessment image feature and a respiratory motion signal, the radiographic data provides an overall range of breath motion amplitudes for all or selected portions of the acquired data. Binned into several amplitude-based bins (with equal counts within each bin) that cover. An amplitude-based respiratory gating is then used in combination with the extracted respiratory motion signal to generate a series of amplitude-based gatings based on LOR (for PET imaging) or projection (for SPECT imaging) timestamps. Consecutive time interval bins are grouped based on their time stamps for image reconstruction into respiration-gated images.

The advantage of this approach is that the displacement versus time curve or the respiratory signal curve can be extracted directly from radial list mode data acquired at short time intervals without time-consuming image reconstruction. This approach is computationally efficient and the statistical noise variation of the extracted data is smaller than that extracted from the reconstructed image. As a result, displacement versus time curves extracted using this approach will be more precise and accurate. Exceptionally, when the patient dataset has fewer counts, using larger time intervals for bins can reduce noise, but at the expense of time accuracy. There is also a practical limitation that time interval bins cannot be larger than a reasonable percentage of a breath.

In this approach, a reconstructed image is used to determine a motion assessment volume over a 3D group of voxels containing features of a candidate image. The motion assessment volume is preferably chosen to contain mostly moving features, with minimal contribution from non-moving features or background to obtain maximum signal-to-noise ratio. Determining the optimal motion assessment volume in an efficient, semi-automatic manner with minimal user intervention is an important feature of this approach, which is described below. In one embodiment, the motion assessment volume is selected using an automated process.

In some embodiments, the motion assessment volume is not selected to precisely register features of the motion assessment image. Indeed, in some embodiments, none of the boundaries of the motion assessment volume are selected to coincide with any boundaries of the features of the motion assessment image. In this approach, for each time interval bin, a displacement versus time curve is generated by an operation that includes calculating the displacement metric of the LOR or projection that is binned into that time interval bin and intersects the motion assessment volume. be. The displacement metric per time interval bin is a statistical metric that serves as a proxy for determining the actual displacement of the motion assessment image features within each time interval bin.

In some embodiments, the motion assessment volume is not selected to precisely register features of the motion assessment image. Indeed, in one embodiment, the motion assessment volume is selected to encompass features of the motion assessment image over the entire cycle of respiratory motion. In this approach, a motion assessment volume is determined from reconstructed images using all LORs or projections from the acquired radiographic data. For each time interval bin, a displacement versus time curve is generated by an operation that includes calculating the displacement metric of the LOR or projection that is binned into that time interval bin and intersects the motion assessment volume. The displacement metric per time interval bin is a statistical metric that serves as a proxy for determining the actual displacement of the motion assessment image features within each time interval bin.

In an exemplary embodiment, the motion assessment volume is selected to include all motion assessment image features during a respiratory motion cycle within the data acquisition period. In another embodiment, the motion assessment volume is selected to partially overlap the features of the motion assessment image, such that the features (or part thereof) of the motion assessment image correspond to the breath of the imaged subject. Move in and out of the motion assessment volume when doing Where the radiopharmaceutical is preferentially accumulating in motion-assessment image features, in the former embodiment, the statistical metric that tracks the centroid location of the total counts within the motion-assessment volume tracks respiration. A curve of displacement metric versus time is provided. In the latter embodiment, the total count within the motion assessment volume increases when some of the features of the motion assessment image move into the motion assessment volume, and the features of the motion assessment image move into the motion assessment volume. Moving out will decrease, thus providing a displacement metric versus time curve that moves periodically with the respiratory cycle.

Other disclosed embodiments include a method for identifying and determining in an efficient manner a motion-assessment volume that either fully encompasses the features of the motion-assessment image or partially overlaps the features of the motion-assessment image. . Here, the maximum intensity gradient is exploited in combination with an average intensity that satisfies some minimum threshold to identify features of the motion assessment image from the reconstructed image. A preferred criterion allows extraction of the largest motion signal with minimal background for the highest possible signal-to-background ratio by minimizing the contribution of counts from stationary parts of the image. It is to determine the motion evaluation volume.

Referring to FIG. 1, an exemplary radiation imaging system comprises a combined positron emission tomography (PET)/transmission computed tomography (CT) imaging device 8 with a patient mounted on a common patient table 14. , both the PET imaging gantry or scanner 10 and the CT gantry or scanner 12 mounted with coaxial bores so as to be mounted within either the CT gantry 12 for CT imaging or the PET gantry 10 for PET imaging. Prepare. The PET imaging gantry or scanner 10 is equipped with a radiation detector that detects 511 keV gamma rays, and LOR is defined by two near-simultaneous gamma-ray detections presumed to result from a single positron-electron annihilation event. In one embodiment, the radiation detector of the PET gantry is a high speed detector capable of detecting the time difference between detections of two 511 keV gamma rays emitted by a single positron-electron annihilation event. This measured time difference allows for further time-of-flight (TOF) localization of positron-electron annihilation events along the LOR. Each LOR is timestamped with the time of acquisition (usually the finite TOF difference is in picoseconds and can be ignored for LOR timestamping purposes). CT gantry 12, if provided, for example, by suitably converting the Hounsfield numbers of CT images 16 to corresponding absorption values at 511 keV (energy of gamma rays emitted during positron-electron annihilation events) A transmission CT image 16 is acquired, which is used to generate the attenuation map 18 . As a non-limiting illustrative example, an exemplary PET/CT imaging device imaging scanner 8 is manufactured by Koninklijke Philips N.V., Eindhoven, The Netherlands. V. The PET gantry of the Vereos™ Digital PET/CT Scanner available from .

An exemplary radiographic imager is a PET imager 10 that acquires radiographic data in the form of a time-stamped LOR. In another embodiment, the radiation imager is a gamma camera that acquires radiation imaging data in the form of single photon emission computed tomography (SPECT) projection data. In SPECT imaging, each projection is defined by a single radiation photon detection event, also time-stamped. As is known in the art, SPECT imaging projections are most commonly acquired by one of two large field-of-view gamma cameras or 2D planar radiation detectors mounted on a rotating gantry. . The 2D projection image consisted of many elongated holes separated by diaphragms made of highly radiation-absorbing atomic weight (high Z) materials such as lead or lead compounds mounted on the head of a radiation detector. , is formed by using a collimator. A collimator is designed to allow radiation photons traveling in a straight line or narrow-angle cone perpendicular to the detection plane to reach the radiation detector, forming a 2D projection image. By rotating the gamma camera around the patient, multiple 2D projections are acquired from different viewpoints and used for image reconstruction of the final 3D or multi-slice SPECT image. Similar to PET imaging, the optional CT gantry 12 can produce CT images 16 that are transformed into attenuation maps 18 for performing attenuation correction during SPECT reconstruction.

As shown schematically in FIG. 1, an electronic processor 20 processes radiographic imaging data 22 acquired by a PET imaging gantry or scanner 10 (including LOR in an exemplary PET imaging embodiment, or alternative SPECT imaging implementations). morphology, including projections acquired by the gamma camera), estimate respiratory motion, and generate reconstructed images. Electronic processor 20 may, for example, store instructions on one or more non-transitory storage media (eg, one or more hard drives, optical disks, solid state drives, or other electronic digital storage devices, various thereof). A computer 24 (eg, a desktop computer, a network-based server computer, a dedicated imager control computer, various combinations thereof, etc.) that executes instructions read from it. Computer 24 typically includes or has operable access to at least one display 26 (e.g., LCD display, plasma display, etc.) for displaying reconstructed images, and optionally, exemplary keyboard 28, exemplary One or more user input devices, such as a trackpad 29 (or mouse, trackball, touch-sensitive overlay on display 26, or other pointing device) are also provided.

Emission imaging data 22 is captured over a relatively long period of time, i.e., a large number of breaths taken by the imaging subject, to provide sufficient emission imaging data to achieve an acceptable signal-to-noise ratio (SNR). captured over a time interval containing As shown schematically in FIG. 1, electronic processor 20, by instructions stored in a non-transitory storage medium, performs a respiratory motion estimation process 30 that operates only on radiographic imaging data 22. programmed. The exemplary respiratory motion estimation process 30 includes an act or subprocess 32 that selects a feature of the motion assessment image and an act or subprocess 34 that computes a displacement versus time curve representing an estimate of the respiratory motion. and

Electronic processor 20 further performs respiratory gating image reconstruction process 40 operating on radiographic data 22, respiratory motion estimation (i.e., displacement versus time curves) output by process 30, and optionally , further programmed by instructions stored in a (same or different) non-transitory storage medium to perform attenuation correction of the reconstructed image on the attenuation map 18 . The exemplary respiratory gating image reconstruction process 40 reconstructs a subset of the radiographic imaging data 22 corresponding to a selected respiratory stage (typically, this stage is stationary and long-lasting, end-expiratory). , operate to produce reconstructed images with reduced respiratory motion blur.

The exemplary electronic processor 20 further performs image display processing 42 to display images displayed on at least the display 26 and printed by a printing device stored in a Picture Archiving and Communication System (PACS). reconstruction, such as a single two-dimensional (2D) slice image, a 2D maximum intensity projection (MIP), a three-dimensional (3D) rendering of a volumetric reconstructed image, etc., and/or otherwise utilized It is programmed by instructions stored in a (same or different) non-transitory storage medium to perform image visualization.

With continued reference to FIG. 1 and further reference to FIG. 2, an exemplary embodiment of the process 32 for selecting features of a motion assessment image is described. In operation 48 the radiographic data 22 is reconstructed to produce a reconstructed image 50 . In general, all radiographic data 22 is reconstructed, thus forming a reconstructed image 50 blurred by respiratory motion. The purpose of the reconstructed image 50 is not to provide a clinical image for medical diagnosis or other uses, but simply to provide a way of roughly delineating features of the image. Accordingly, the image reconstruction operation 48 does not use attenuation correction using the attenuation map 18, but optionally uses other techniques, such as using larger voxel sizes than are used in reconstructing clinical images. , use non-iterative reconstruction such as filtered backprojection, and/or otherwise to achieve fast and computationally efficient reconstruction. (In an exemplary embodiment, reconstructed image 50 is a single image reconstructed from all radiographic data 22, but alternatively, from all radiographic data, or selected portions thereof, , it is conceivable to generate a reconstructed image 50 comprising a series of reconstructed images or a reconstructed volume.) In operation 52, the reconstructed image 50 is used to identify features of the motion assessment image. In some embodiments, a single pre-selected motion assessment image feature is selected. For example, if the radiopharmaceutical is known to preferentially accumulate in the myocardium, operation 52 tracks the centroid of the total counts within a motion assessment volume encompassing the entire myocardium during the respiratory motion cycle. . In another example, if high radiopharmaceutical uptake was found in the liver, pattern matching would be used to identify the upper portion of the liver to determine the features of the motion assessment image.

However, in the exemplary embodiment of FIG. 2, the goal is the motion assessment image regardless of the type of imaging performed, the type of radiopharmaceutical used, details of the particular imaging subject, and/or other. is to provide a more general process for selecting features of Thus, the exemplary selection operation 52 selects the optimal image from a set of candidate anatomical features 54, e.g., heart, liver, and lungs (or lung/thoracic diaphragm boundary) in the illustrative example. operates to select features of Each of these anatomical features may represent, as non-limiting examples, average image intensity or total counts within a motion assessment volume that wholly or partially encompasses candidate image features within reconstructed image 50. , or combinations thereof, using count statistics selection criteria 56 . The features of each candidate image are evaluated based on selection criteria 56, and the image features best suited for body motion estimation evaluated by selection criteria 56 are selected as body motion evaluation image features 60. FIG. For example, this selection 52 selects the candidate image feature with the highest average image intensity. This is because the mean image intensity represents the substantial data acquired for the candidate image features, which makes the estimation of respiratory motion more accurate. As another example, selection 52 selects the optimal motion assessment volume for extracting the highest motion signal against the background from the candidate image features. The feature extraction method appropriately utilizes the maximum intensity gradient in combination with the average intensity satisfying some minimum threshold. This criterion yields the highest motion signal-to-background ratio by minimizing the contribution of counts from non-moving parts of the image.

With continued reference to FIG. 1 and further reference to FIG. 3, an exemplary embodiment of a process 34 for generating a displacement versus time curve that enables respiratory motion assessment is described. In operation 62, the radiographic imaging data 22 is binned into time interval bins based on LOR or projection timestamps. The bin time interval improves the temporal resolution (using a shorter time interval) versus the signal-to-noise ratio (SNR, using a longer time interval improves by having more data per bin). are selected to balance The time intervals should be small enough so that each time interval is a relatively small fraction of a breath. The goal is to generate a displacement value for each time interval bin, thereby creating a displacement versus time curve.

A simple approach is to reconstruct the radiographic data for each time interval bin, generate a corresponding "bin image", and directly measure the displacement of the motion assessment image feature 60 within each bin image. In this case, the displacement metric is not a statistical displacement metric, but rather a direct measure of displacement. However, such "binned images" suffer from a generally high noise level. The average adult, at rest, takes about 12-20 breaths per minute, which equates to one breath every 3-5 seconds. As such, the duration of the time interval bins should generally be less than 1 second. With this in mind, the amount of data per time interval bin is relatively small and the reconstructed "bin image" per time interval bin is low signal and high noise.

Exemplary process 34 is advantageous because it does not need to perform any computationally expensive image reconstruction operations, and is therefore computationally efficient, and does not need to process low SNR "binned images." . In an exemplary process, at operation 64, a motion assessment volume 66 is defined that overlaps the features 60 of the motion assessment image. A motion assessment volume 66 spans a group of voxels to yield a reasonably high signal value. The motion assessment volume 66 is generally not selected to precisely align with the features 60 of the motion assessment image. Indeed, in some embodiments, none of the boundaries of the motion assessment volume are selected to coincide with any boundaries of the features of the motion assessment image. Thus, the motion assessment volume 66 can be defined to be computationally efficient and not require complex surface delineation or feature segmentation processes. For example, in one approach, the motion assessment volume 66 is selected as a standard three-dimensional (3D) shape, such as a cylinder or cube, placed over the features 60 of the motion assessment image. Optionally, the standard shape is scaled based on the size of the features 60 of the motion assessment image to define a motion assessment volume 66 . The motion assessment volume 66 only partially overlaps (and preferably only partially overlaps in some embodiments) the features of the motion assessment image. A motion estimation volume 66 is defined, and for each time interval bin, the statistical displacement of the LOR or projection that is both (i) binned into the time interval bin and (ii) intersected with the motion estimation volume 66. An act 68 of calculating a metric is performed. Such a statistical displacement metric for successive time interval bins then produces a displacement metric versus time curve 70 that serves as a proxy for determining the actual displacement of the motion assessment image features within each time interval bin. Form.

4 and 5, an illustrative example of process 34 shown in FIG. 3 is described. FIG. 4 illustrates a motion assessment image feature 60 including a boundary 72 between a lung 74 and a thoracic diaphragm 76 . Since features 60 in the motion assessment image are identified in the respiratory motion blurred reconstructed image 50, the lung/diaphragm boundary 72 is a blurred boundary, as shown schematically by shading and hatching in FIG. is. In this example, the radiopharmaceutical is assumed to preferentially accumulate in the lungs 74 compared to the thoracic diaphragm 76, which is shown schematically in FIG. Note that it shows An exemplary motion assessment volume 66 is shown in dashed lines in FIG. 4 and is a cubic or spherical volume that overlaps the lung/diaphragm boundary 72 . FIG. 5 schematically illustrates operation 68 of FIG. FIG. 5 shows the results for successive time interval bins, labeled "Time Bin n", "Time Bin n+1", "Time Bin n+2", "Time Bin n+3", "Time Bin n+4", and "Time Bin n+5". is plotted. As shown at the top of FIG. 5, "time bin n" and "time bin n+4" are two consecutive end-expiratory times, while "time bin n+2" is the intervening end-inspiratory time.

The top row of FIG. 5 is shown in the noise-free state (which is physically infeasible as the total count in each time interval bin would be low, resulting in a low SNR binned image). ), showing an idealized “bin image” that visualizes what would be reconstructed for each time interval bin. A motion assessment volume 66 is superimposed on each of these virtual idealized bin images. Note that the motion assessment volume 66 is stationary, ie at the same position in space for each virtual bin image. In this example, the displacement metric calculated for each time interval bin is the total count of LORs or projections that belong to the time interval bin and intersect the motion assessment volume 66 . As seen in schematic FIG. 5, at end-expiration (time interval bins n and n+4) the diaphragm 76 is fully relaxed, resulting in the lungs contracting to their minimum volume and thus only a small portion of the lungs 74 ( or none at all) extends into the motion assessment volume. Since the lungs 74 contain a higher concentration of radiopharmaceutical than the diaphragm 76, this results in lower total counts (total count metric is a proxy for the displacement of the lung/diaphragm boundary, hence this curve is also referred to herein as the displacement metric versus time curve 70). As the imaged subject begins to inhale, the thoracic diaphragm 76 contracts, thereby causing the lungs 74 to inflate, as seen in the virtual idealized binned images of time bin intervals n+1, n+5. Stretch. This causes a larger portion of the lungs 74 to move into the motion assessment volume 66 , increasing the total count of the motion assessment volume 66 and increasing the total count of the metric as seen in the displacement metric vs. time curve 70 . rises. At end-inspiration (time interval bin n+2), the lungs 74 are maximally extended into the motion assessment volume 66 to the peak of the displacement metric versus time curve 70 . Later, as the imaged subject begins to exhale (time interval bin n+3), the expansion of the lungs 74 into the motion assessment volume 66 decreases and the displacement metric versus time curve 70 reaches a minimum at end-expiration. up to the descending part.

Note that FIG. 5 is a schematic diagram and, in particular, does not accurately show the actual duration of the various respiratory stages. It will also be appreciated that in a real resting adult, the end-expiratory phase is the longest phase, typically about 40% of the total respiratory cycle.

Referring to FIG. 6, there is shown another approach for defining the motion assessment volume 66, where the displacement metric is the center of mass (COM) of the "diagonalized" volume within 66. It is well suited for performing operation 68. This example also uses a motion assessment image feature 60 that includes a (respiratory motion blurred) boundary 72 between a lung 74 and a thoracic diaphragm 76 . Again, it is assumed that the radiopharmaceutical preferentially accumulates in the lungs 74 compared to the thoracic diaphragm 76, again schematically illustrated in FIG. shown in In this embodiment, the motion assessment volume 66 is defined as follows. A motion-blurred boundary 72 is identified in the reconstructed image 50 . This blurred boundary 72 is identified, for example, as an area of image intensity gradient. The identified blurred boundary 72 is then enlarged by some selected amount (eg, a fixed percentage) to form a motion assessment volume 66, shown schematically in dashed lines in FIG. In this illustrative example, the displacement metric is the center of gravity (COM) 78 of the volume containing the "diagonal lines" shown in FIG. COM 78 is calculated, for example, by taking the center point (x _i , y _i , z _i ) of each LOR or projection (indexed by i) and averaging the center points according to: x _COM =Σ _i x _i , y _COM = Σ _i y _i , and z _COM = Σ _i z _i . However, assuming a prone or supine patient, where the z-direction is defined as the patient's axial anatomical direction, the lungs expand "downward" along the axial or z-direction during inspiration and is expected to contract “up” along the axial or z-direction, so only the axial component z _COM =Σ _i z _i is of primary interest and can serve as the COM value 78 . The COM-based approach is particularly effective when the concentration of radiopharmaceutical in the volume containing the "diagonal line" is much higher than in the lung 74 and thoracic diaphragm 76 volumes.

Illustrative examples make use of the lung/thoracic diaphragm boundary, but more generally (i) indicates a gradient or boundary over which the counts vary, and (ii) moves in correlation with the respiratory cycle. , any organ or tissue can be used. For example, a motion-assessed image feature may be the liver or heart, in which radiopharmaceuticals used in radiometric imaging preferentially accumulate, and the motion-assessed volume correlates the heart or liver with the respiratory cycle. It is suitably selected to overlap the heart or liver to enter and exit the motion assessment volume. Note that this approach can be advantageously adapted to situations where the features 60 of the motion assessment image are only partially within the field of view (FOV) of the radiographic imaging. This is addressed by selecting a motion assessment volume 66 that overlaps and intersects the boundaries of the motion assessment image features 60 that remain completely within the FOV throughout the respiratory cycle.

In a deformation technique, the motion assessment volume 66 may be selected to completely overlap the heart, liver, or other motion assessment image feature 60 with some space around the edges. By using the center of gravity (COM) as a statistical displacement metric, the anterior-posterior motion of the heart or liver within the motion assessment volume 66 can be tracked in correlation with the respiratory cycle. In this deformation approach, the motion assessment volume does not intersect the boundaries of the features of the motion assessment image, rather the motion assessment volume completely encompasses the features of the motion assessment image.

In the example of FIG. 5, the total count is highest at end inspiration (time bin n+2) and lowest at end expiration (time bins n and n+4). This is because the higher count lungs 74 enter the motion assessment volume 66 when the imaged subject inhales and exit the motion assessment volume 66 when the imaged subject exhales. Due to the shape. Another selection of the motion assessment image features 60 and the motion assessment volume 66 is between the respiratory stages (e.g., end-inspiration and end-expiration) and the statistical displacement metric calculated for the motion assessment volume 66. , other relations are generated. This relationship is easily determined for a given geometry by considering an idealized virtual bin image, as shown in FIG. 5 for an exemplary lung/diaphragm boundary.

In general, the use of motion assessment volumes 66 selected to overlap motion assessment image features 60 optimized to exhibit large gradients and/or concentrations of the radiopharmaceutical is particularly sensitive to respiratory motion. By selectively processing the radiation imaging data contributing to the motion assessment volume 66 selected to be, the SNR can be effectively increased. A larger motion assessment volume 66 results in more counts and increases the signal. On the other hand, selecting the motion assessment volume to exclude significant portions of the reconstructed image 50 that are not respiration-variant reduces noise and increases overall SNR. Furthermore, computationally expensive "binned image" reconstruction is avoided by calculating a statistical displacement metric for the radiographic data within the motion assessment volume. The selection of statistical displacement metrics can also be optimized to maximize sensitivity to respiratory motion. For example, in the case where the motion-assessment volume 66 completely encompasses the motion-assessment image features 60, the displacement metric of total counts is that the motion-assessment image features 60 remain within the motion-assessment volume 66 throughout the respiratory cycle. If it stays full, it is a bad choice because the total count does not change. On the other hand, the COM displacement metric is effective because it tracks the forward and backward motion of the motion assessment image feature 60 even when the motion assessment image feature 60 remains entirely within the motion assessment volume 66 . be.

Yet another alternative embodiment contemplates identifying the motion assessment volume 66 from the reconstructed image 50 without identifying any particular motion assessment image features 60 . For example, the motion assessment volume 66 is selected to be a cylindrical volume or other standard volume approximately centered on the subject's torso. This approach takes advantage of the fact that the central torso region contains or is adjacent to the lungs and diaphragm and is therefore likely to exhibit substantial motion that correlates with the respiratory cycle. In this case, the lung, diaphragm, or lung/diaphragm boundary need not be specifically identified by segmentation, contouring, or other computationally complex processing.

For PET imaging with time-of-flight (TOF) localization along the LOR, this TOF information is used to refine the displacement metric used in operation 68 . For TOF-PET, the statistical displacement metric is the weighting of each LOR intersecting the motion evaluation volume 66 by the probability that the LOR occurred within the motion evaluation volume 66, determined from the TOF localization of the LOR. is calculated using This probability is determined based on how much of the TOF probability distribution along the LOR is within the motion assessment volume 66, and for TOF-PET image reconstruction, a given LOR with TOF localization is It is analogous to estimating the probability of occurrence at a particular voxel.

Illustrative examples of suitable motion assessment volumes 66 are each a single concatenated volume. However, it is conceivable that the motion assessment volume 66 includes two (or more) spatially separated volumetric regions. Stated another way, the motion assessment volume 66 includes more than one constituent volume, and in operation 68 the metric may differ from volume to volume. In such a case, the motion assessment volume 66 includes two or more constituent volumes, and the displacement vs. time curve 70 is obtained by calculating a statistical displacement metric for each constituent volume for each time interval bin, and configuring generated by an operation that involves combining (eg, adding together or averaging) statistical displacement metrics for the volume to be applied.

Referring again to FIG. 1 and with additional reference to FIG. 7, an exemplary embodiment of a respiratory gated image reconstruction process 40 is described. At operation 82, a radiographic data subset is selected from radiographic data 22 corresponding to a particular respiratory stage (or more generally, gating interval). Typically, the respiratory phase selected for reconstruction is end-expiratory since it is the longest resting phase of the respiratory cycle, although another phase can be selected. The selected subset of radiographic data generally includes a number of time interval bins from successive respiratory cycles, e.g., in the example of FIG. Note that we include (ie, combine) the radiographic imaging data in these time bins because they correspond. In practice, due to the length of the physiological end-expiratory phase, several consecutive time interval bins are likely to correspond to each end-expiratory period and thus be synthesized in operation 82 . In operation 84, stage-specific radiographic data subsets are reconstructed. The image reconstruction operation 84 is intended to produce high quality reconstructed images suitable for tasks such as clinical diagnosis or clinical evaluation. Accordingly, the image reconstruction operation 84 preferably performs attenuation correction using the attenuation map 18, preferably using an image reconstruction algorithm that is expected to suppress image artifacts or other image defects. do. For example, image reconstruction 84 may use iterative image reconstruction techniques such as maximum likelihood expectation-maximization (ML-EM), ordered-subset expectation-maximization (OS-EM), and the like. optionally includes regularization using edge-preserving noise-suppressing priors, scatter correction, or other known techniques to improve image quality. The resulting reduced respiratory motion artifact (e.g., reduced respiratory motion blur) reconstructed image 86 is preferably processed by image display processing 42 and displayed on display 26. (See Figure 1).

Referring again to FIG. 1 , the respiratory stage-specific image reconstruction 40 described with reference to FIG. 7 is one exemplary application of respiratory motion assessment 30 . Other applications are contemplated, for example, the displacement vs. time curve 70 is a motion correction that spatially shifts the LOR or projection of the radiographic imaging data 22 based on the range of displacements indicated by the displacement vs. time curve 70. Used as input to the algorithm. In another possible application, if the subject is instructed to hold their breath during imaging, the displacement vs. time curve 70 is used to detect the time intervals during which the subject did not comply with this instruction. be done. Any imaging data identified by the displacement versus time curve 70 acquired while the subject was not holding their breath is then discarded. These are merely illustrative examples.

Although described in terms of estimating respiratory motion, the disclosed techniques can be applied to motion due to the cardiac cycle, or known and common volitional motion, such as turning the head from side to side (e.g., for imaging the brain). It can also be used to estimate other types of body motion, such as movement. In the case of the cardiac cycle, features 60 in the motion assessment image are, for example, the heart, aorta, or other major arteries. For brain imaging, the motion assessment image feature 60 is, for example, the brain/skull boundary.

The invention has been described with reference to preferred embodiments. Modifications and alterations may occur to others upon reading and understanding the preceding detailed description. It is intended that all such modifications and variations be included in the invention insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims (15)

an electronic processor;
a non-transitory storage medium readable and executable by said electronic processor for storing instructions for performing a respiratory body motion estimation method, said respiratory body motion estimation method comprising: teeth,
Reconstructing radiographic data to produce a reconstructed image by using voxel sizes larger than those used to reconstruct clinical images , said radiographic data being positron emission tomography (PET) ) generating LORs (coincidence lines) acquired by an imaging device or projections acquired by a gamma camera;
determining a motion assessment volume over a group of voxels using the reconstructed image;
binning the radiographic data into time interval bins based on timestamps of the LOR or the projection;
For each of said time interval bins, an operation comprising computing a statistical displacement metric of said LORs or said projections binned into said time interval bins and intersecting said motion assessment volume produces a curve of displacement versus time. generating. 2. The radiographic imaging data processing apparatus of claim 1, wherein generating the displacement versus time curve does not include performing any image reconstruction operations to provide a clinical image . 3. A radiographic imaging data processing apparatus according to claim 1 or 2, wherein said statistical displacement metric is a total count of said LORs or said projections binned into said time interval bins and intersecting said motion assessment volume. 3. A radiographic imaging data processing apparatus according to claim 1 or 2, wherein the statistical displacement metric is the center of gravity of the LOR or projections binned into the time interval bins and intersecting the motion assessment volume. The radiographic imaging data has a LOR with time-of-flight (TOF) localization acquired by a TOF-PET imager, and calculating the statistical displacement metric comprises the TOF localization of the LOR. 5. Radiographic data processing according to claim 3 or 4, comprising weighting each LOR intersecting said motion assessment volume by the probability that said LOR occurred within said motion assessment volume, determined from Device. Determining the body motion assessment volume using the reconstructed image includes:
identifying features of body motion assessment images in the reconstructed images;
selecting the body motion assessment volume that includes features of the body motion assessment image or overlaps features of the body motion assessment image;
6. A radiographic imaging data processing apparatus according to any one of claims 1 to 5, wherein no boundary of said motion assessment volume is selected to coincide with any boundary of a feature of said motion assessment image. a feature of the motion assessment image is the lungs, the motion assessment volume is selected to intersect or at least partially encompass a boundary between the lungs and the thoracic diaphragm;
7. The radiographic imaging data processing apparatus of claim 6, wherein the motion assessment volume is selected by enlarging the boundary between the lungs and the thoracic diaphragm. Identifying features of the body motion evaluation image includes:
identifying features of a plurality of candidate images within the reconstructed image;
calculating a statistical measure for each feature of the candidate image;
and selecting the features of the body motion evaluation image from a plurality of the candidate image features based on the statistical measures calculated for the candidate image features. A radiographic data processing apparatus as described. The statistical measure for each of the candidate image features is the sum of the average image intensity of the candidate image feature in the reconstructed image and the maximum image intensity gradient of the candidate image feature. 9. A radiographic imaging data processing apparatus according to claim 8, comprising at least one. 10. Apparatus according to any one of claims 1 to 9, wherein reconstructing the radiographic data for generating the reconstructed image does not use an attenuation map for correcting attenuation. . The non-transitory storage medium further stores instructions readable and executable by the electronic processor for performing a process of reconstructing a gated image, the process comprising:
using the displacement vs. time curve generated by the respiratory motion estimation method and the timestamp of the LOR or the projection to select a subset of the radiographic data corresponding to a respiratory gating interval. and,
reconstructing the subset of the radiographic data corresponding to the respiratory gating interval to generate a reconstructed image corresponding to the respiratory gating interval. 10. A radiation imaging data processing apparatus according to 1. Reconstructing the subset of the radiographic imaging data corresponding to the respiratory gating interval to generate the reconstructed image corresponding to the respiratory gating interval performs attenuation correction using an attenuation map. 12. A radiographic imaging data processing apparatus according to claim 11, comprising: a positron emission tomography (PET) imager or gamma camera for acquiring radiographic data;
A radiation imaging data processing device according to claim 11 or 12;
A radiographic imaging apparatus comprising: a display operably connected to the radiographic data processing apparatus for displaying reconstructed images corresponding to respiratory gating intervals. said motion assessment volume comprising two or more constituent volumes, said displacement vs. time curve calculating a statistical displacement metric for each of said constituent volumes for each time interval bin; 13. A radiographic imaging data processing apparatus according to any one of claims 1 to 12, wherein the statistical displacement metric is generated by an operation comprising: A non-transitory storage medium readable and executable by an electronic processor for storing instructions for performing a respiratory body motion estimation method by processing operations, the processing operations comprising:
Reconstructing radiographic data to produce a reconstructed image by using voxel sizes larger than those used to reconstruct clinical images , said radiographic data being positron emission tomography (PET) ) generating LORs (coincidence lines) acquired by an imaging device or projections acquired by a gamma camera;
identifying features of a plurality of candidate images within the reconstructed image;
calculating a statistical measure for each feature of the candidate image;
selecting a body motion assessment image feature from the plurality of candidate image features based on the calculated statistical measure for the candidate image feature;
generating from the radiographic imaging data a displacement versus time curve representing motion of features of the body motion assessment image.

Images